This is a span-based system for Question Answering (QA) in Chinese. Potential uses include online automated customer services, etc.
We use Anaconda3 as default. You can create a conda environment by environment.yml.
P.S. torch==1.9.0+cu111 might need to be installed from the official website.
Create a sub-directory named ckpt/ and put epoch=9-step=14210.ckpt inside it. The .ckpt file is available on 百度网盘(提取码:49w0).
First, launch the application
flask run --host=0.0.0.0
Then, query the system via a POST request. For example, in Python, run the following lines:
>>> resp = requests.post("http://localhost:5000/predict", files={"context_path": open("context.txt", "r", encoding="utf-8"), "question_path": open("question.txt", "r", encoding="utf-8")})
>>> resp.json()
{'answer': '1995年，率先在业内提出"六保"服务--保真、保质、保价、保换、保修、保洗，用优良的品质、优质的服务营造"**钻石专家"的品牌形象。', 'question': '金伯利的
六项保障服务是什么？'}